The impact of MEMS (Microelectromechanical Systems) devices on the microelectronics industry has been extremely significant in recent years. Indeed, MEMS is one of the fastest growing areas of microelectronics. The growth of MEMS has been enabled, to a large extent, by the extension of silicon-based photolithography to the manufacture of micro-scale mechanical devices and structures. Photolithographic techniques, of course, rely on reliable etching techniques, which allow accurate etching of a silicon substrate revealed beneath a mask.
MEMS devices have found applications in a wide variety of fields, such as in physical, chemical and biological sensing devices. One important application of MEMS devices is in inkjet printheads, where micro-scale actuators for inkjet nozzles may be manufactured using MEMS techniques. The present Applicant has developed printheads incorporating MEMS ink ejection devices and these are the subject of a number of patents and patent applications listed in the Cross References section above and all of which are incorporated herein by reference.
Typically a MEMS inkjet printhead (“MEMJET printhead”) is comprised of a plurality of integrated circuits, with each integrated circuit having several thousand nozzles. Each nozzle comprises an actuator for ejecting ink, which may be, for example, a thermal bend actuator (e.g. U.S. Pat. No. 6,322,195) or a bubble-forming heater element actuator (e.g. U.S. Pat. No. 6,672,709). The integrated circuits are manufactured using MEMS techniques, meaning that a high nozzle density and, hence, high resolution printheads can be mass-produced at relatively low cost.
In the manufacture of MEMS printhead integrated circuits, it is often required to perform deep or ultradeep etches. Etch depths of about 3 □m to 10 □m may be termed “deep etches”, whereas etch depths of more than about 10 □m may be termed “ultradeep etches.
MEMS printhead integrated circuits typically require delivery of ink to each nozzle through individual ink supply channels having a diameter of about 20 □m. These ink channels are typically etched through wafers having a thickness of about 200 □m, and therefore place considerable demands on the etching method employed. It is especially important that each ink channel is perpendicular to the wafer surface and does not contain kinks, sidewall projections (e.g. grassing) or angular junctions, which can interfere with the flow of ink.
In the Applicant's U.S. patent application Ser. No. 10/728,784 (Applicant Ref: MTB08) and Ser. No. 10/728,970 (Applicant Ref: MTB07), both of which are incorporated herein by reference, there is described a method of fabricating inkjet printheads from a wafer 5 having a drop ejection side and an ink supply side. Referring to FIG. 1, there is shown a typical MEMS nozzle arrangement 1 comprising a bubble-forming heater element actuator assembly 2. The actuator assembly 2 is formed in a nozzle chamber 3 on the passivation layer 4 of a silicon wafer 5. The wafer typically has a thickness “B” of about 200 □m, whilst the nozzle chamber typically occupies a thickness “A” of about 20 □m. The nozzle chamber 3 has an inlet 8, which joins an ink supply channel (not shown in FIG. 1) in the silicon wafer 5.
Referring to FIG. 2, an ink supply channel 6 is formed in the wafer 5 by first etching a trench through the CMOS metallization layer of a TEOS interconnect 7 and partially through the wafer 5 from the ink ejection side 20 of the wafer. Once formed, the trench is plugged with photoresist 10 whilst nozzle structures are formed on the ink ejection side 20 of the wafer. After formation of the nozzle arrangement 1, the ink supply channel 6 is formed by ultradeep etching from the ink supply side 30 of the wafer to and past the photresist plug 10.
Referring to FIG. 3, the photoresist plug 10 is finally stripped away to form the inlet 8. The inlet 8 provides fluid connection between the ink supply channel 6 and the nozzle chamber 3. FIGS. 2 and 3 also shows the CMOS drive circuitry 9, which is provided between the wafer 5 and the interconnect 7.
The “back-etching” of the ink supply channel avoids filling and removing an entire 200 □m long ink supply channel with resist whilst nozzle structures in the wafer are being lithographically formed. However, there are a number of problems associated with back-etching the ink supply channels in this way. Firstly, the mask on the ink supply side needs to be carefully aligned so that the etched channels meet the trenches plugged with photoresist, and do not damage the drive circuitry 9. Secondly, the etching needs to be perpendicular and anisotropic to a depth of about 200 □m.
Several methods for etching ultradeep trenches into silicon are known in the art. All these methods involve deep reactive ion etching (DRIE) using a gas plasma. The semiconductor substrate, with a suitable mask disposed thereon, is placed on a lower electrode in a plasma reactor, and exposed to an ionized gas plasma formed from a mixture of gases. The ionized plasma gases (usually positively charged) are accelerated towards the substrate by a biasing voltage applied to the electrode. The plasma gases etch the substrate either by physical bombardment, chemical reaction or a combination of both. Etching of silicon is usually ultimately achieved by formation of volatile silicon halides, such as SiF4, which are carried away from the etch front by a light inert carrier gas, such as helium.
Anisotropic etching is usually achieved by depositing a passivation layer onto the base and sidewalls of the trench as it is being formed, and selectively etching the base of the trench using the gas plasma.
The most widely used method for achieving ultradeep anisotropic etching is the “Bosch process”, described in U.S. Pat. Nos. 5,501,893 and 6,284,148. This method involves alternating polymer deposition and etching steps. After formation of a shallow trench, a first polymer deposition step deposits a polymer onto the base and sidewalls of the trench. The polymer is deposited by a gas plasma formed from a fluorocarbon gas (e.g CHF3, C4F8 or C2F4) in the presence or in the absence of an inert gas. In the subsequent etching step, the plasma gas mix is changed to SF6/Ar. The polymer deposited on the base of the trench is quickly broken up by ion assistance in the etching step, while the sidewalls remain protected. Hence, anisotropic etching may be achieved.
However, a major disadvantage of the Bosch process is that polymer deposition and etching steps need to be alternated, which means continuously alternating the gas composition of the plasma, thereby slowing etch rates. This alternation, in turn, leads to uneven trench sidewalls, characterized by scalloped surface formations.
A further disadvantage of the Bosch process is that it leaves a hydrophobic fluorocarbon polymer coating (‘Bosch polymer’) on the trench sidewalls. In ink channels for inkjet printheads, it is desirable to provide hydrophilic sidewalls so that ink is drawn into ink supply channels by a capillary action.
Hitherto, such polymeric coatings were removed by either O2 ashing or a ‘wet’ clean. The standard industry method for removing ‘Bosch polymer’ residues is to use an EKC™ wet clean, followed by DI rinse(s) and spin drying. However, both O2 ashing and EKC™ cleans suffer from serious problems during MEMS processing of a silicon wafer. Usually, a silicon wafer being processed in a MEMS foundry is attached to a handle wafer, such as a glass handle wafer, using a release tape such as Revalpha™ thermal release tape. The handle wafer is necessary so that the silicon wafer can be handled during backside processing steps, without damaging any sensitive structures (e.g. inkjet nozzles) already fabricated the frontside of the wafer. As described above, typically, during MEMS manufacture of an inkjet printhead, nozzles are first formed on a frontside of wafer and then ink supply channels are etched from a backside of the wafer. Before performing backside processing steps (e.g. wafer grinding, etching), the frontside of the wafer, having a protective photoresist coating, is usually attached to a handle wafer using a thermal release tape. The thermal release tape comprises a film having a thermal release adhesive on one side. The thermal release adhesive conveniently allows the silicon wafer to be detached from the handle wafer by controlled heating after the completion of backside processing steps.
However, standard O2 ashing ovens are run out about 220-240° C., which is well above the release temperature of standard thermal release tape (about 160-180° C.). Consequently, the silicon wafer delaminates from the handle wafer during O2 ashing in a standard ashing oven. Lowering the temperature of ashing oven produces unacceptably low ashing rates and does not ensure complete removal of any polymeric residues coated on sidewalls of etched features.
Moreover, standard EKC™ cleans tend to attack thermal release tape chemically, also causing highly undesirable delamination.
An additional problem with ultradeep anisotropic etching is that anisotropy tends to be lost when the etch front meets the photoresist plug 10. FIGS. 2 and 3 show an idealized fabrication process in which the etch front continues to etch anistropically and flush against the photoresist plug 10 when the etch front meets the photoresist. In practice however, and referring to FIG. 4, when the etch front meets the photoresist plug 10, the etch flares radially outwards, leaving a spiked circumferential rim 13 around the photoresist plug. Radial flaring where the etch front meets the photoresist plug is believed to be due to mutual charge repulsion between a charge built up on the photoresist and the charged ions in the plasma. The radially flared etch terminus 12 and the corresponding spiked circumferential rim 13 are shown in FIG. 4.
The radially flared etch terminus 12 is undesirable in the final printhead integrated circuit when the photoresist is removed (FIG. 5). The flared etch terminus 12 acts as a pocket, which can trap slow-moving ink or gas bubbles. This can lead to disruption of the flow of ink from the ink channel 6 to the inlet 8 and the nozzle downstream.
Moreover, the spiked rim 13, which faces the flow of oncoming ink, is relatively weakly supported by the bulk wafer 5 and can easily break off into one or more slivers. The creation of slivers in the ink flow is highly undesirable and will typically result in failure of the nozzle downstream.
It would be desirable to provide a process, which exposes angular or spiked surface features in an etched trench, thereby facilitating modification of such features. It would further be desirable to provide a process for manufacturing a printhead integrated circuit, whereby ink channels formed in the process have improved surface profiles.
It would be also be desirable to provide an alternative process for removing a hydrophobic polymeric layer deposited on the sidewalls of a trench during etching. It would further be desirable to provide a process for removing such polymers, which is compatible with industry standard thermal release tape, and does not cause delamination. It would further be desirable to provide a process for manufacturing a printhead integrated circuit, whereby ink channels formed in the process have improved surface properties.